# RxNExtract
A professional-grade system for extracting chemical reaction information from procedure texts using fine-tuned LLM with Dynamic prompting and self grounding.
[](https://badge.fury.io/py/rxnextract)
[](https://opensource.org/licenses/MIT)
[](https://www.python.org/downloads/)
[](https://hub.docker.com/r/chemplusx/rxnextract)
[](https://huggingface.co/chemplusx/rxnextract)
## โจ Key Features
- **Advanced AI**: Fine-tuned LLM with dynamic prompting and self-grounding
- **Modular Architecture**: Clean, maintainable codebase with separation of concerns
- **Multiple Interfaces**: CLI, interactive mode, batch processing, and programmatic API
- **Memory Efficient**: 4-bit quantization support for deployment on various hardware
- **Comprehensive Analysis**: Error analysis, ablation studies, statistical testing, and uncertainty quantification
- **Easy Installation**: One-command installation via PyPI, Conda, or Docker
## ๐ Quick Start
### 30-Second Demo
```bash
# Install
pip install rxnextract
# Use
python -c "
from chemistry_llm import ChemistryReactionExtractor
extractor = ChemistryReactionExtractor.from_pretrained('chemplusx/rxnextract-complete')
procedure = 'Add 5g NaCl to 100mL water and stir for 30 minutes at room temperature.'
results = extractor.analyze_procedure(procedure)
print('Reactants:', results['extracted_data']['reactants'])
print('Conditions:', results['extracted_data']['conditions'])
"
```
### Try Without Installation
[](https://colab.research.google.com/github/chemplusx/RxNExtract/blob/master/examples/quickstart.ipynb)
[](https://huggingface.co/spaces/chemplusx/rxnextract-demo)
## ๐ฆ Installation Options
### Option 1: PyPI (Recommended)
```bash
pip install rxnextract # Basic installation
pip install rxnextract[gpu] # GPU support
pip install rxnextract[full] # All features
```
### Option 2: Conda
```bash
conda install -c conda-forge rxnextract
```
### Option 3: Docker
```bash
docker pull chemplusx/rxnextract:latest
docker run -it --gpus all chemplusx/rxnextract:latest
```
### Option 4: From Source
```bash
git clone https://github.com/chemplusx/RxNExtract.git
cd RxNExtract
pip install -e .
```
## ๐ฏ Performance Highlights
Our complete framework achieves significant improvements over baseline methods:
| Metric | Baseline | RxNExtract | Improvement |
|--------|----------|------------|-------------|
| **Complete Reaction Accuracy** | 23.4% | **52.1%** | **+122.6%** |
| **Entity F1 Score** | 0.674 | **0.856** | **+27.0%** |
| **Role Classification Accuracy** | 68.2% | **85.9%** | **+25.9%** |
| **Condition F1 Score** | 0.421 | **0.689** | **+63.7%** |
**Error Reduction**: 47.8-55.2% across all major error categories
**Statistical Significance**: McNemar's ฯยฒ = 134.67 (p < 0.001), Cohen's d = 0.82
## ๐ Documentation
| Document | Description |
|----------|-------------|
| **[Installation & Setup Guide](docs/INSTALLATION.md)** | Detailed installation instructions, system requirements, and configuration |
| **[Usage Guide & Examples](docs/USAGE.md)** | Comprehensive usage examples, API reference, and advanced features |
| **[Analysis & Evaluation](docs/ANALYSIS.md)** | Complete analysis framework, metrics, and research reproducibility |
| **[Changelog](docs/CHANGELOG.md)** | Version history and release notes |
## ๐ฌ Research Applications
Perfect for:
- **Chemical Literature Mining**: Extract structured reaction data from papers
- **Procedure Standardization**: Convert natural language to structured formats
- **Database Curation**: Automated reaction database construction
- **Educational Tools**: Teaching reaction analysis and extraction
- **Research Reproducibility**: Systematic evaluation of extraction methods
## ๐ค Community & Support
### Getting Help
- ๐ **Documentation**: [docs.rxnextract.org](https://docs.rxnextract.org)
- ๐ **Bug Reports**: [GitHub Issues](https://github.com/chemplusx/RxNExtract/issues)
- ๐ฌ **Discussions**: [GitHub Discussions](https://github.com/chemplusx/RxNExtract/discussions)
- ๐ง **Email**: support@rxnextract.org
### For Experimental Chemists
- ๐ฏ **One-click installations** via PyPI and Conda
- ๐ณ **Docker containers** for consistent environments
- ๐ **User-friendly tutorials** and examples
- ๐ **Video tutorials** and webinars
### For Developers
- ๐ง **Extensive API documentation**
- ๐งช **Comprehensive test suite**
- ๐๏ธ **Modular architecture** for easy extension
- ๐ **Contributing guidelines** and code standards
## ๐ Quick Examples
### Basic Usage
```python
from chemistry_llm import ChemistryReactionExtractor
# Initialize extractor
extractor = ChemistryReactionExtractor.from_pretrained("chemplusx/rxnextract-complete")
# Analyze procedure
procedure = """
Dissolve 5.0 g of benzoic acid in 100 mL of hot water.
Add 10 mL of concentrated HCl and cool the solution.
Filter the precipitated product and wash with cold water.
"""
results = extractor.analyze_procedure(procedure)
print(results['extracted_data'])
```
### Command Line Interface
```bash
# Interactive mode
rxnextract --interactive
# Batch processing
rxnextract --input procedures.txt --output results.json
# Single procedure
rxnextract --procedure "Add 2g NaCl to 50mL water"
```
### Analysis & Research
```python
from chemistry_llm.analysis import ErrorAnalyzer, AblationStudy
# Error analysis
analyzer = ErrorAnalyzer()
error_results = analyzer.analyze_prediction_errors(predictions, ground_truth)
# Ablation study
ablation = AblationStudy(model_path="./model")
study_results = ablation.run_complete_study(test_data, ground_truth)
```
## ๐๏ธ System Requirements
| Component | Minimum | Recommended |
|-----------|---------|-------------|
| **Python** | 3.8+ | 3.9+ |
| **RAM** | 8GB | 16GB+ |
| **GPU Memory** | 4GB | 12GB+ |
| **Storage** | 20GB | 50GB+ |
| **CPU** | 4 cores | 8+ cores |
*Note: Requirements are for inference only. Fine-tuning requires additional resources.*
## ๐ Data and Software Availability
**Code Repository**: All code used in this study is available under the MIT License at https://github.com/chemplusx/RxNExtract. The MIT License permits unrestricted use, modification, and distribution, making it suitable for both academic research and commercial applications.
**Pre-trained Models**:
- HuggingFace Hub: [chemplusx/rxnextract-complete](https://huggingface.co/chemplusx/rxnextract-complete)
- Model cards with training details, performance metrics, and usage guidelines
**Package Distribution**:
- PyPI: `pip install rxnextract`
- Conda-Forge: `conda install -c conda-forge rxnextract`
- Docker Hub: `docker pull chemplusx/rxnextract:latest`
**Datasets**: Training and evaluation datasets are available at [Zenodo DOI: 10.5281/zenodo.XXXXXX](link-to-zenodo)
**Reproducibility**: Complete analysis scripts and configuration files are provided to reproduce all results presented in the paper.
## ๐ค Contributing
We welcome contributions from the community! Whether you're fixing bugs, adding features, improving documentation, or sharing use cases, your help makes RxNExtract better.
### Quick Contributing Guide
1. Fork the repository
2. Create a feature branch (`git checkout -b feature/amazing-feature`)
3. Make your changes and add tests
4. Ensure all tests pass (`python -m pytest`)
5. Submit a pull request
See our [Contributing Guidelines](CONTRIBUTING.md) for detailed instructions.
## ๐ License & Citation
**License**: This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for complete terms.
**Citation**: If you use RxNExtract in your research, please cite our paper:
```bibtex
@article{rxnextract2025,
title={RxNExtract: A Professional-Grade System for Chemical Reaction Extraction using Fine-tuned LLMs},
author={[Your Authors]},
journal={[Journal Name]},
year={2025},
doi={[DOI]}
}
```
## ๐ Links
- **Homepage**: https://github.com/chemplusx/RxNExtract
- **Documentation**: https://docs.rxnextract.org
- **PyPI Package**: https://pypi.org/project/rxnextract/
- **Docker Images**: https://hub.docker.com/r/chemplusx/rxnextract
- **HuggingFace Models**: https://huggingface.co/chemplusx/rxnextract-complete
- **Paper**: [Link to published paper]
---
Raw data
{
"_id": null,
"home_page": null,
"name": "rxnextract",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.8",
"maintainer_email": null,
"keywords": "chemistry, reaction, extraction, llm, nlp, machine-learning",
"author": null,
"author_email": "ChemPlusX <xyzmadhavi@gmail.com>",
"download_url": "https://files.pythonhosted.org/packages/80/9b/c92a784567a4fa1c32d9edb7a7101436f21ea80dd85eab434e699575e1e1/rxnextract-1.2.5.tar.gz",
"platform": null,
"description": "# RxNExtract\r\n\r\nA professional-grade system for extracting chemical reaction information from procedure texts using fine-tuned LLM with Dynamic prompting and self grounding.\r\n\r\n[](https://badge.fury.io/py/rxnextract)\r\n[](https://opensource.org/licenses/MIT)\r\n[](https://www.python.org/downloads/)\r\n[](https://hub.docker.com/r/chemplusx/rxnextract)\r\n[](https://huggingface.co/chemplusx/rxnextract)\r\n\r\n## \u2728 Key Features\r\n\r\n- **Advanced AI**: Fine-tuned LLM with dynamic prompting and self-grounding\r\n- **Modular Architecture**: Clean, maintainable codebase with separation of concerns\r\n- **Multiple Interfaces**: CLI, interactive mode, batch processing, and programmatic API\r\n- **Memory Efficient**: 4-bit quantization support for deployment on various hardware\r\n- **Comprehensive Analysis**: Error analysis, ablation studies, statistical testing, and uncertainty quantification\r\n- **Easy Installation**: One-command installation via PyPI, Conda, or Docker\r\n\r\n## \ud83d\ude80 Quick Start\r\n\r\n### 30-Second Demo\r\n\r\n```bash\r\n# Install\r\npip install rxnextract\r\n\r\n# Use\r\npython -c \"\r\nfrom chemistry_llm import ChemistryReactionExtractor\r\nextractor = ChemistryReactionExtractor.from_pretrained('chemplusx/rxnextract-complete')\r\nprocedure = 'Add 5g NaCl to 100mL water and stir for 30 minutes at room temperature.'\r\nresults = extractor.analyze_procedure(procedure)\r\nprint('Reactants:', results['extracted_data']['reactants'])\r\nprint('Conditions:', results['extracted_data']['conditions'])\r\n\"\r\n```\r\n\r\n### Try Without Installation\r\n[](https://colab.research.google.com/github/chemplusx/RxNExtract/blob/master/examples/quickstart.ipynb)\r\n[](https://huggingface.co/spaces/chemplusx/rxnextract-demo)\r\n\r\n## \ud83d\udce6 Installation Options\r\n\r\n### Option 1: PyPI (Recommended)\r\n```bash\r\npip install rxnextract # Basic installation\r\npip install rxnextract[gpu] # GPU support\r\npip install rxnextract[full] # All features\r\n```\r\n\r\n### Option 2: Conda\r\n```bash\r\nconda install -c conda-forge rxnextract\r\n```\r\n\r\n### Option 3: Docker\r\n```bash\r\ndocker pull chemplusx/rxnextract:latest\r\ndocker run -it --gpus all chemplusx/rxnextract:latest\r\n```\r\n\r\n### Option 4: From Source\r\n```bash\r\ngit clone https://github.com/chemplusx/RxNExtract.git\r\ncd RxNExtract\r\npip install -e .\r\n```\r\n\r\n## \ud83c\udfaf Performance Highlights\r\n\r\nOur complete framework achieves significant improvements over baseline methods:\r\n\r\n| Metric | Baseline | RxNExtract | Improvement |\r\n|--------|----------|------------|-------------|\r\n| **Complete Reaction Accuracy** | 23.4% | **52.1%** | **+122.6%** |\r\n| **Entity F1 Score** | 0.674 | **0.856** | **+27.0%** |\r\n| **Role Classification Accuracy** | 68.2% | **85.9%** | **+25.9%** |\r\n| **Condition F1 Score** | 0.421 | **0.689** | **+63.7%** |\r\n\r\n**Error Reduction**: 47.8-55.2% across all major error categories\r\n**Statistical Significance**: McNemar's \u03c7\u00b2 = 134.67 (p < 0.001), Cohen's d = 0.82\r\n\r\n## \ud83d\udcda Documentation\r\n\r\n| Document | Description |\r\n|----------|-------------|\r\n| **[Installation & Setup Guide](docs/INSTALLATION.md)** | Detailed installation instructions, system requirements, and configuration |\r\n| **[Usage Guide & Examples](docs/USAGE.md)** | Comprehensive usage examples, API reference, and advanced features |\r\n| **[Analysis & Evaluation](docs/ANALYSIS.md)** | Complete analysis framework, metrics, and research reproducibility |\r\n| **[Changelog](docs/CHANGELOG.md)** | Version history and release notes |\r\n\r\n## \ud83d\udd2c Research Applications\r\n\r\nPerfect for:\r\n- **Chemical Literature Mining**: Extract structured reaction data from papers\r\n- **Procedure Standardization**: Convert natural language to structured formats\r\n- **Database Curation**: Automated reaction database construction\r\n- **Educational Tools**: Teaching reaction analysis and extraction\r\n- **Research Reproducibility**: Systematic evaluation of extraction methods\r\n\r\n## \ud83e\udd1d Community & Support\r\n\r\n### Getting Help\r\n- \ud83d\udcda **Documentation**: [docs.rxnextract.org](https://docs.rxnextract.org)\r\n- \ud83d\udc1b **Bug Reports**: [GitHub Issues](https://github.com/chemplusx/RxNExtract/issues)\r\n- \ud83d\udcac **Discussions**: [GitHub Discussions](https://github.com/chemplusx/RxNExtract/discussions)\r\n- \ud83d\udce7 **Email**: support@rxnextract.org\r\n\r\n### For Experimental Chemists\r\n- \ud83c\udfaf **One-click installations** via PyPI and Conda\r\n- \ud83d\udc33 **Docker containers** for consistent environments\r\n- \ud83d\udcd6 **User-friendly tutorials** and examples\r\n- \ud83c\udf93 **Video tutorials** and webinars\r\n\r\n### For Developers\r\n- \ud83d\udd27 **Extensive API documentation**\r\n- \ud83e\uddea **Comprehensive test suite**\r\n- \ud83c\udfd7\ufe0f **Modular architecture** for easy extension\r\n- \ud83d\udccb **Contributing guidelines** and code standards\r\n\r\n## \ud83d\udd11 Quick Examples\r\n\r\n### Basic Usage\r\n```python\r\nfrom chemistry_llm import ChemistryReactionExtractor\r\n\r\n# Initialize extractor\r\nextractor = ChemistryReactionExtractor.from_pretrained(\"chemplusx/rxnextract-complete\")\r\n\r\n# Analyze procedure\r\nprocedure = \"\"\"\r\nDissolve 5.0 g of benzoic acid in 100 mL of hot water.\r\nAdd 10 mL of concentrated HCl and cool the solution.\r\nFilter the precipitated product and wash with cold water.\r\n\"\"\"\r\n\r\nresults = extractor.analyze_procedure(procedure)\r\nprint(results['extracted_data'])\r\n```\r\n\r\n### Command Line Interface\r\n```bash\r\n# Interactive mode\r\nrxnextract --interactive\r\n\r\n# Batch processing\r\nrxnextract --input procedures.txt --output results.json\r\n\r\n# Single procedure\r\nrxnextract --procedure \"Add 2g NaCl to 50mL water\"\r\n```\r\n\r\n### Analysis & Research\r\n```python\r\nfrom chemistry_llm.analysis import ErrorAnalyzer, AblationStudy\r\n\r\n# Error analysis\r\nanalyzer = ErrorAnalyzer()\r\nerror_results = analyzer.analyze_prediction_errors(predictions, ground_truth)\r\n\r\n# Ablation study\r\nablation = AblationStudy(model_path=\"./model\")\r\nstudy_results = ablation.run_complete_study(test_data, ground_truth)\r\n```\r\n\r\n## \ud83c\udfd7\ufe0f System Requirements\r\n\r\n| Component | Minimum | Recommended |\r\n|-----------|---------|-------------|\r\n| **Python** | 3.8+ | 3.9+ |\r\n| **RAM** | 8GB | 16GB+ |\r\n| **GPU Memory** | 4GB | 12GB+ |\r\n| **Storage** | 20GB | 50GB+ |\r\n| **CPU** | 4 cores | 8+ cores |\r\n\r\n*Note: Requirements are for inference only. Fine-tuning requires additional resources.*\r\n\r\n## \ud83d\udcca Data and Software Availability\r\n\r\n**Code Repository**: All code used in this study is available under the MIT License at https://github.com/chemplusx/RxNExtract. The MIT License permits unrestricted use, modification, and distribution, making it suitable for both academic research and commercial applications.\r\n\r\n**Pre-trained Models**: \r\n- HuggingFace Hub: [chemplusx/rxnextract-complete](https://huggingface.co/chemplusx/rxnextract-complete)\r\n- Model cards with training details, performance metrics, and usage guidelines\r\n\r\n**Package Distribution**:\r\n- PyPI: `pip install rxnextract` \r\n- Conda-Forge: `conda install -c conda-forge rxnextract`\r\n- Docker Hub: `docker pull chemplusx/rxnextract:latest`\r\n\r\n**Datasets**: Training and evaluation datasets are available at [Zenodo DOI: 10.5281/zenodo.XXXXXX](link-to-zenodo)\r\n\r\n**Reproducibility**: Complete analysis scripts and configuration files are provided to reproduce all results presented in the paper.\r\n\r\n## \ud83e\udd1d Contributing\r\n\r\nWe welcome contributions from the community! Whether you're fixing bugs, adding features, improving documentation, or sharing use cases, your help makes RxNExtract better.\r\n\r\n### Quick Contributing Guide\r\n1. Fork the repository\r\n2. Create a feature branch (`git checkout -b feature/amazing-feature`)\r\n3. Make your changes and add tests\r\n4. Ensure all tests pass (`python -m pytest`)\r\n5. Submit a pull request\r\n\r\nSee our [Contributing Guidelines](CONTRIBUTING.md) for detailed instructions.\r\n\r\n## \ud83d\udcc4 License & Citation\r\n\r\n**License**: This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for complete terms.\r\n\r\n**Citation**: If you use RxNExtract in your research, please cite our paper:\r\n```bibtex\r\n@article{rxnextract2025,\r\n title={RxNExtract: A Professional-Grade System for Chemical Reaction Extraction using Fine-tuned LLMs},\r\n author={[Your Authors]},\r\n journal={[Journal Name]},\r\n year={2025},\r\n doi={[DOI]}\r\n}\r\n```\r\n\r\n## \ud83d\udd17 Links\r\n\r\n- **Homepage**: https://github.com/chemplusx/RxNExtract\r\n- **Documentation**: https://docs.rxnextract.org\r\n- **PyPI Package**: https://pypi.org/project/rxnextract/\r\n- **Docker Images**: https://hub.docker.com/r/chemplusx/rxnextract\r\n- **HuggingFace Models**: https://huggingface.co/chemplusx/rxnextract-complete\r\n- **Paper**: [Link to published paper]\r\n\r\n---\r\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "Professional chemistry reaction extraction using fine-tuned LLMs",
"version": "1.2.5",
"project_urls": {
"Analysis": "https://chemplusx.github.io/RxNExtract/ANALYSIS/",
"Bug Reports": "https://github.com/chemplusx/RxNExtract/issues",
"Changelog": "https://chemplusx.github.io/RxNExtract/CHANGELOG/",
"Documentation": "https://chemplusx.github.io/RxNExtract/",
"Homepage": "https://github.com/chemplusx/RxNExtract",
"Installation Guide": "https://chemplusx.github.io/RxNExtract/INSTALLATION/",
"Source": "https://github.com/chemplusx/RxNExtract",
"Usage Guide": "https://chemplusx.github.io/RxNExtract/USAGE/"
},
"split_keywords": [
"chemistry",
" reaction",
" extraction",
" llm",
" nlp",
" machine-learning"
],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "87f115431d7e8b17c20db99355a6da50caf93b734b94eef4d8ca2e522fba98e9",
"md5": "85d75ed710cd15a57bb6b61defc9bd5a",
"sha256": "4dc345c10e2281c66c185b832e178b31756951a8471b654d15733be13d6fd476"
},
"downloads": -1,
"filename": "rxnextract-1.2.5-py3-none-any.whl",
"has_sig": false,
"md5_digest": "85d75ed710cd15a57bb6b61defc9bd5a",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.8",
"size": 50077,
"upload_time": "2025-09-04T21:39:41",
"upload_time_iso_8601": "2025-09-04T21:39:41.293901Z",
"url": "https://files.pythonhosted.org/packages/87/f1/15431d7e8b17c20db99355a6da50caf93b734b94eef4d8ca2e522fba98e9/rxnextract-1.2.5-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "809bc92a784567a4fa1c32d9edb7a7101436f21ea80dd85eab434e699575e1e1",
"md5": "c4e14ae7587ea00948e1bce41f6224dc",
"sha256": "a78252090f49b34434a0938d65adb5bf926da8cee1af0fda77dce2ce6b483fe6"
},
"downloads": -1,
"filename": "rxnextract-1.2.5.tar.gz",
"has_sig": false,
"md5_digest": "c4e14ae7587ea00948e1bce41f6224dc",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.8",
"size": 48788,
"upload_time": "2025-09-04T21:39:42",
"upload_time_iso_8601": "2025-09-04T21:39:42.823211Z",
"url": "https://files.pythonhosted.org/packages/80/9b/c92a784567a4fa1c32d9edb7a7101436f21ea80dd85eab434e699575e1e1/rxnextract-1.2.5.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-09-04 21:39:42",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "chemplusx",
"github_project": "RxNExtract",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"requirements": [
{
"name": "torch",
"specs": [
[
">=",
"2.0.0"
]
]
},
{
"name": "transformers",
"specs": [
[
">=",
"4.35.0"
]
]
},
{
"name": "peft",
"specs": [
[
">=",
"0.7.0"
]
]
},
{
"name": "bitsandbytes",
"specs": [
[
">=",
"0.41.0"
]
]
},
{
"name": "accelerate",
"specs": [
[
">=",
"0.24.0"
]
]
},
{
"name": "numpy",
"specs": [
[
">=",
"1.21.0"
]
]
},
{
"name": "pandas",
"specs": [
[
">=",
"1.3.0"
]
]
},
{
"name": "pyyaml",
"specs": [
[
">=",
"6.0"
]
]
},
{
"name": "tqdm",
"specs": [
[
">=",
"4.64.0"
]
]
},
{
"name": "click",
"specs": [
[
">=",
"8.0.0"
]
]
},
{
"name": "pytest",
"specs": [
[
">=",
"7.0.0"
]
]
},
{
"name": "pytest-cov",
"specs": [
[
">=",
"4.0.0"
]
]
},
{
"name": "black",
"specs": [
[
">=",
"23.0.0"
]
]
},
{
"name": "isort",
"specs": [
[
">=",
"5.12.0"
]
]
},
{
"name": "flake8",
"specs": [
[
">=",
"6.0.0"
]
]
},
{
"name": "mypy",
"specs": [
[
">=",
"1.0.0"
]
]
},
{
"name": "statsmodels",
"specs": [
[
"==",
"0.14.5"
]
]
},
{
"name": "flash-attn",
"specs": [
[
">=",
"2.0.0"
]
]
},
{
"name": "wandb",
"specs": [
[
">=",
"0.15.0"
]
]
}
],
"lcname": "rxnextract"
}
JSON
